Artificial Neural Network Predictive Modelling for floor tiling production rate.
Kanyike, Carlton Daniel
MetadataShow full item record
Demand for tiling in the construction industry is at an all-time high and consequently, costs of labour on construction sites have exponentially increased. This only becomes profitable at an optimal production rate. Unfortunately, poor data mining, and analysis practices in the construction industry have made it next to impossible to establish benchmarks against which optimal production rate can be measured. This is supported by substandard quantification of the various factors affecting construction project productivity and/or production rate as well as a wanting understanding of the level of relevance that each of these factors has on the eventual outcome of a construction project. Through studying the factors affecting the production rate of floor tiling in the Ugandan context, measuring the duration of the floor tiling operation, and formulating an Artificial Neural Network predictive model, this study provides a solution to the above-mentioned issues. Data was collected by way of an interview survey, observation and on-site measurement and was analysed using MATLAB, EasyFit, SPSS and Microsoft Excel Softwares. The Artificial Neural Network was trained, tested and validated using the data collected as a basis and thereafter run using unseen data. The relationship between the predictions and the variables was diagnosed using a correlation matrix and an R value of 0.8852 was returned showing a very successful model. Similarly, the performance of the model was tracked basing on the mean square of differences between the predictions and the targets. The mean square error of the model used was 0.05134 indicating less than a tenth of error.